90% of brands invisible in AI search results

What happened

Per Search Engine Journal, a new study from SEO agency Victorious found that 90% of brands receive zero mentions across AI search engines. The research examined the relationship between traditional organic rankings and citations inside generative answer engines, and the gap is stark: a small minority of domains capture nearly all of the visibility that LLMs surface to users.

Search Engine Journal reports that the study identified four insights tying classical SEO performance to AI mention frequency. The headline finding is the concentration: the vast majority of brands that compete in Google's blue links are simply absent when ChatGPT, Perplexity, Google AI Overviews, or Gemini generate an answer. Being on page one is no longer a proxy for being in the answer.

The second-order finding matters more for strategy. Traditional ranking signals still correlate with AI mentions, but the correlation is loose enough that high-ranking domains are routinely skipped, and lower-ranking domains with the right content shape are routinely cited. The mapping is not one-to-one.

Why it matters for your brand

If you run marketing at a global bank, an industrial group, a UN agency, or a foundation, the 90% figure is the number you should be putting in front of your CEO this quarter. It reframes the visibility problem. For two decades, the question was "where do we rank?" The new question is "are we in the answer at all?" For nine out of ten brands, the answer today is no.

The implication for financial services is sharp. Wealth managers, custodians, and asset managers have spent years optimising thought leadership for Google. That work produces PDFs and insight pages that rank, but the studies suggest those assets are not being cited when a CFO asks ChatGPT about counterparty risk or treasury strategy. The competitive set inside an LLM answer is narrower than the competitive set on a SERP, and exclusion compounds: if a model has never seen your brand associated with a topic, it will not start citing you because you published one more whitepaper.

For multilaterals and policy institutions, the stakes are different but no less acute. UNDRR, the World Bank's CGAP, the IMF, the OECD: these organisations are sources of record. They expect to be cited. If their research is not surfacing in generative answers about climate risk, financial inclusion, or sovereign debt, the policy conversation moves on without them, and private-sector consultancies fill the vacuum. Authority that does not appear in the answer is authority that does not exist for a growing share of decision-makers.

Industrial groups face the brand-substitution problem. When a procurement lead asks Gemini about low-carbon cement or grid-scale storage suppliers, the brands the model names become the shortlist. HOLCIM, Siemens Energy, Schneider: these companies are not competing for clicks, they are competing to be one of three names in a paragraph. If you are in the 90% with zero mentions for your category's key prompts, you are not on the shortlist, and you will not know it until a deal you should have been in goes to someone else.

For philanthropic and policy institutions, the funding implication is direct. Programme officers and grantees increasingly use LLMs for landscape scans. The foundations and think tanks that get named in those scans shape the agenda. The ones that do not, do not. Content strategy now has to be measured by citation share inside model outputs, not just by downloads or media pickup.

The signal in context

The 90% number lines up with what other measurement firms have been reporting through 2024 and 2025: AI answer engines concentrate citations on a narrow band of sources, weight Wikipedia and Reddit heavily, and reward content that is structured to answer a specific question rather than rank for a broad keyword. Traditional SEO performance is a necessary but not sufficient condition for AI visibility. The brands winning citations are doing three things in combination: they are publishing content that maps cleanly to prompt intent, they are being discussed on third-party platforms the models trust, and they have enough domain authority that the model treats them as a safe attribution.

What the Victorious study adds is a quantified gap. Nine in ten is not a long tail problem; it is the median. For CMOs, that means AI visibility is still an open competitive field. The brands that move now, with measurement, structured content, and earned third-party presence, will define the citation set for their categories before the models harden their preferences. The window is not closed. But the assumption that good SEO will carry you into the answer engine is finished.